3dfx vs NVIDIA: how the Voodoo empire collapsed

In the late 1990s, before graphics cards became the size of roof tiles and required enough electricity to make your energy supplier send a thank-you card, there was 3dfx. For PC gamers of that era, the company’s Voodoo cards were not just components; they were status symbols, magic tricks, and tiny green passports to the future. You did not simply install a Voodoo card. You performed a ritual involving drivers, cables, IRQ anxiety, and the sincere hope that Windows would not punish you for your ambition. For a few dazzling years, 3dfx Interactive stood at the center of the PC gaming revolution. Its technology made games smoother, faster, richer, and more convincing than anything most home users had seen before. The company helped turn the graphics card from a dull necessity into the beating heart of a gaming machine. Then, almost as quickly as it rose, it collapsed. By the end of 2000, 3dfx was selling most of its assets to NVIDIA, the rival that had learned to play the game faster, leaner, and with fewer sentimental attachments to the past. However, the story of 3dfx is not a simple tale of failure…. 

The beginning

3dfx was founded in 1994 by Ross Smith, Scott Sellers, and Gary Tarolli, three engineers with roots in Silicon Graphics, a company associated with the kind of high-end 3D work that ordinary PC owners could only admire from a safe financial distance. Silicon Graphics machines were used for serious visualization, film effects, simulation, and professional 3D graphics. They were powerful, expensive, and about as likely to appear in a teenager’s bedroom as a helicopter.

The founders of 3dfx saw an opportunity. The personal computer was changing. Processors were getting faster, memory was becoming more affordable, and games were beginning to demand more than simple 2D sprites and clever tricks. Doom had shown that players wanted speed and atmosphere. Descent had pushed players into true 3D spaces. Quake was about to make every existing PC feel old, tired, and personally insulted.

At first, 3dfx looked toward arcade machines, where dedicated graphics hardware already made sense. Arcades were still ahead of home computers, and players were used to paying for better visuals one coin at a time. But the home PC market was catching up quickly. The question was no longer whether ordinary players wanted 3D acceleration. The question was who would give it to them first in a way that felt essential. 3dfx’s answer was Voodoo Graphics.

The first Voodoo cards were not complete graphics cards in the modern sense. They handled 3D acceleration only, meaning users still needed a separate 2D graphics card. The Voodoo card sat alongside it and connected through a VGA pass-through cable, a setup that looked slightly improvised even when everything was working properly. But once a supported game launched, the awkwardness disappeared. Textures became smoother. Frame rates improved. Worlds gained depth and atmosphere. Suddenly, PC gaming felt less like a clever illusion and more like a door opening.

The moment everything changed

The arrival of Voodoo Graphics was one of those moments when a technology does not merely improve something; it changes what people expect from it. Before Voodoo, many PC games used software rendering, relying on the main processor to draw 3D worlds. Developers were talented enough to make this work, but there were limits. Games could be fast, or detailed, or smooth, but rarely all three at once. Voodoo shifted that balance.

For gamers, the change was immediate and emotional. A familiar game running through a Voodoo card could look like a different product. The jagged, muddy, flickering world of early 3D suddenly became cleaner and more fluid. Explosions had more punch. Corridors had more atmosphere. Racing games felt faster. Shooters felt more physical. Even the menus somehow seemed more serious, as if your PC had started wearing sunglasses indoors.

This was the genius of 3dfx. The company did not ask consumers to care about abstract engineering. It showed them something they could see instantly. You did not need a white paper to understand Voodoo. You needed five minutes with Quake.

That made 3dfx different from many technology companies. Its advantage was technical, but its impact was emotional. Gamers did not talk about fill rates and texture units because they were preparing for an exam. They talked about them because those numbers translated into bragging rights, smoother play, and the deeply satisfying feeling that your machine was better than your friend’s machine.

The success

3dfx’s rise was powered by more than hardware. The company understood that graphics technology lives or dies by software support, so it created Glide, its own graphics API. At a time when PC 3D standards were still messy, Glide gave developers a fast, reliable way to target Voodoo hardware. It was one of the company’s smartest moves.

Glide helped create a powerful loop. Developers supported Glide because Voodoo cards were popular. Gamers bought Voodoo cards because games supported Glide. Reviewers praised Voodoo because the best games looked fantastic on it. Retailers liked Voodoo because people actually walked in asking for it by name. For a while, the whole thing fed itself beautifully, like capitalism with texture filtering.

The Voodoo brand quickly became associated with premium PC gaming. Cards based on 3dfx chips appeared from major board manufacturers, which helped put the technology into more hands without forcing 3dfx to build every card itself. That partner model mattered. It meant 3dfx could focus on the chips and the developer ecosystem while companies with retail experience handled much of the board business.

Then came Voodoo2, and 3dfx’s reputation reached another level. Voodoo2 was faster, more capable, and even more desirable than the original. Most famously, it supported SLI, short for Scan-Line Interleave, which allowed two cards to work together. This was absurd, wonderful, expensive, and extremely on-brand for PC gaming. One card was impressive. Two cards suggested you had either excellent priorities or no adult supervision.

For enthusiasts, Voodoo2 SLI became the dream setup. It was the kind of thing people discussed in forums, magazines, computer shops, and school corridors with the intensity normally reserved for sports cars or secret government aircraft. 3dfx had not just captured a market. It had captured imaginations.

The culture of voodoo

The 3dfx logo before a game became a small event. It told players that what followed had been blessed by the hardware gods. The company’s name carried a sense of mystery and power that most competitors lacked. “Voodoo” sounded exciting. It sounded dangerous. It sounded like something your motherboard might not fully approve of, which only made it better.

This cultural power should not be underestimated. In the 1990s PC market, hardware was often confusing, beige, and badly explained. 3dfx gave gamers a simple promise: buy this, and your games will look better. That clarity was priceless. The company did not need to convince players that 3D acceleration was theoretically useful. It made the benefit obvious.

Magazines loved the story because it was visual and dramatic. Screenshots looked better. Benchmarks improved. Developers had something new to discuss. Gamers had something new to desire. The whole industry suddenly had momentum, and 3dfx was riding at the front with one hand on the wheel and the other probably adjusting a BIOS setting.

The company also benefited from perfect timing. PC gaming was entering one of its most experimental and exciting periods. First-person shooters, flight sims, racing games, and action adventures were all pushing toward richer 3D worlds. Every new release seemed to demand better hardware. Every upgrade promised a more convincing illusion. 3dfx arrived just as the market was ready to believe.

What went right

The first thing 3dfx got right was focus. The company did not try to be everything to everyone at the beginning. It built hardware that made 3D games look spectacular, and it targeted the audience most likely to care. That sounds obvious now, but it was a decisive advantage at a time when the consumer 3D market was still taking shape.

The second thing it got right was developer support. Glide was not just a technical tool; it was a business weapon. By making it easier for developers to get strong results on Voodoo hardware, 3dfx encouraged more games to support its cards. That gave consumers confidence. Nobody wants to buy expensive hardware that sits inside a PC doing nothing except warming the case.

The third thing it got right was branding. Voodoo was one of the great product names in PC hardware. It was memorable, emotional, and instantly tied to the experience the cards delivered. Plenty of competitors had technically capable products with names that sounded like printer cartridges. 3dfx had Voodoo. That mattered.

The fourth thing it got right was image quality. Even when rivals began to catch up in performance, many players still liked the way Voodoo cards rendered games. The experience felt smooth and polished. 3dfx understood that graphics were not only about numbers; they were about what players saw and felt during play.

Most of all, 3dfx made graphics hardware exciting. Before Voodoo, a graphics card was often just another component. After Voodoo, it became the star of the system. That shift helped create the modern GPU culture, for better and worse. Without 3dfx, today’s graphics-card obsession might still exist, but it would have arrived with less drama, fewer legends, and probably fewer people arguing on the internet about anisotropic filtering at two in the morning.

The market changes

The problem for 3dfx was that the market it helped create did not stand still. At first, a separate 3D accelerator made sense. Many users already had 2D cards, and the Voodoo board could simply add high-performance 3D when needed. But consumers eventually wanted simpler solutions. They wanted one card that handled both 2D and 3D. They wanted fewer cables, fewer compatibility worries, and fewer reasons to crawl behind a desk with a flashlight in their mouth.

Competitors moved aggressively into that space. NVIDIA, ATI, Matrox, S3, and others all wanted a piece of the growing 3D market. NVIDIA, in particular, became the most dangerous rival because it moved quickly and learned quickly. Its products improved at a rapid pace, and its business model allowed board partners to build and sell cards while NVIDIA focused on chip development and technology roadmaps.

At the same time, the software landscape was shifting. Glide had once been a major strength, but Direct3D and OpenGL were becoming more important. Developers did not want to support endless proprietary paths forever. Broader standards made it easier to reach more hardware. As those standards improved, Glide became less essential.

This was a classic technology trap. A proprietary advantage can help a company rise, but if the wider industry standard catches up, that same advantage can begin to look like isolation. 3dfx had built a castle. Unfortunately, the market was moving toward highways.

The STB decision

In 1998, 3dfx made the decision that many people now see as the beginning of the end: it bought STB Systems, a graphics card manufacturer. The logic was understandable. By owning a board maker, 3dfx could sell its own branded cards, control the full product, improve access to large PC manufacturers, and capture more profit from each sale.

On paper, this probably looked like a bold step toward maturity. In practice, it damaged one of the company’s greatest strengths.

Before the STB deal, 3dfx sold chips to multiple board partners. Those partners helped spread Voodoo cards widely and gave the company a broad retail presence. After buying STB, 3dfx became a direct competitor to many of those partners. The companies that had helped build the Voodoo empire now had a reason to look elsewhere. And elsewhere was waiting.

NVIDIA benefited enormously from this shift. It continued supplying chips to board partners that wanted strong products without competing against their supplier. As 3dfx became more vertically integrated, NVIDIA remained more flexible. In a market where product cycles were getting shorter and competition was getting nastier, flexibility mattered.

The STB acquisition also made 3dfx more complicated internally. It was no longer just a chip company with developer relationships and a powerful brand. It now had to manage board manufacturing, inventory, retail distribution, and OEM sales. That is a lot of extra machinery to bolt onto a company already fighting one of the fastest-moving battles in consumer technology.

There is nothing inherently wrong with vertical integration. Apple has done rather well with it, give or take a few dongles. But timing and execution are everything. For 3dfx, the move came just as the graphics market demanded speed, focus, and close partner relationships. The company chose more control at the exact moment it needed more agility.

The competition catches up

The late 1990s graphics market became a street fight in silicon. NVIDIA’s RIVA TNT and TNT2 put real pressure on 3dfx. Then the GeForce 256 arrived and changed the conversation again. NVIDIA marketed it as the first GPU, pushing hardware transform and lighting as the next major step in PC graphics. Whether every game immediately needed those features was almost beside the point. NVIDIA had seized the future-facing narrative.

3dfx still had loyal customers, but the tone of the market had changed. In the Voodoo1 and Voodoo2 era, 3dfx seemed like the obvious choice for serious gamers. By the Voodoo3 era, it was still competitive, but no longer untouchable. Rivals had caught up in important ways, and some were moving faster in features that reviewers and enthusiasts increasingly cared about.

The Voodoo3 sold well, but it also exposed 3dfx’s growing problem. It was a strong card, but not a revolutionary one. It lacked some features competitors were promoting, including 32-bit color rendering. To many ordinary players, this may not have mattered much in daily use. To reviewers, enthusiasts, and the great online benchmarking court of public opinion, it mattered plenty.

Then came Voodoo5 5500, a card that remains fascinating because it was both impressive and too late. Its full-scene anti-aliasing was genuinely appealing. It made jagged edges smoother and games cleaner. 3dfx was betting on image quality at a time when many people were obsessed with raw speed and next-generation feature lists. In some ways, the company was ahead of the conversation. Today, players understand very well that visual smoothness, stability, and image quality matter.

But the market in 2000 was unforgiving. NVIDIA’s GeForce2 cards were fast, modern, and aggressively positioned. Voodoo5 was a good product that needed to be a market-shaking product. It was charming, capable, and burdened with the terrible responsibility of saving a company that had already lost too much time.

The people behind the rise

The 3dfx story is full of people who shaped not only the company but the wider graphics industry. Ross Smith, Scott Sellers, and Gary Tarolli brought deep technical experience and a clear sense that consumer 3D graphics was about to become enormous. Their work helped take ideas that had lived in expensive workstations and specialized machines and bring them into ordinary homes.

Gordon Campbell helped provide early backing and business support, giving the young company room to pursue its vision. In a hardware startup, money is not just fuel; it is oxygen. Without it, even the best ideas remain trapped in notebooks and lab prototypes.

Greg Ballard, who served as chief executive during the company’s explosive growth period, became associated with the confident version of 3dfx: ambitious, visible, and convinced that it could lead the market on its own terms. During the boom years, that confidence looked justified. Voodoo was everywhere that mattered.

Later, Alex Leupp had the far less enviable task of steering the company during its decline. By then, 3dfx was facing delayed products, falling revenue, fierce competition, and shrinking options. It is easy to romanticize the founders and the glory years. It is harder, but fairer, to remember that the people managing the collapse were dealing with problems that had been building for years.

Outside the company, figures such as John Carmack of id Software played a major role in shaping 3dfx’s reputation. Carmack’s work on games like Quake made high-performance 3D acceleration feel essential. When influential developers supported a platform, gamers paid attention. In that era, developer credibility could move hardware.

And across the battlefield was NVIDIA, led by Jensen Huang. NVIDIA did not win simply because it made one better chip. It won through execution, speed, partnerships, and a ruthless ability to keep pushing the market forward. While 3dfx tried to preserve and extend its Voodoo kingdom, NVIDIA built a machine designed to keep moving.

What went wrong

The fall of 3dfx was not caused by one bad decision, though the STB deal certainly deserves a long, uncomfortable stare. The collapse came from several problems arriving together. The company alienated important board partners, took on manufacturing complexity, fell behind in product timing, leaned too long on Glide, struggled to match NVIDIA’s pace, and failed to secure the kind of OEM presence that could have stabilized its business.

There was also a cultural issue. 3dfx had been the company that defined the rules. That is a wonderful position until the rules change. The company seemed slow to accept that the market no longer revolved around Glide, Voodoo branding, and dedicated enthusiast loyalty. Gamers still loved 3dfx, but love becomes complicated when the other company is offering faster cards and shinier benchmark results.

The graphics industry also became brutally fast. A missed product cycle could do lasting damage. Two missed cycles could be fatal. Consumers wanted constant improvement. Reviewers wanted new features. Developers wanted standards. OEMs wanted reliable supply, competitive pricing, and roadmaps they could trust. Investors wanted growth. In short, everyone wanted everything, preferably yesterday, and with better drivers.

3dfx was trying to solve too many problems while NVIDIA was executing with terrifying focus. By 2000, the gap had become too large. The Voodoo5 could not reverse the story. The GigaPixel acquisition hinted at future technology, but future technology does not help much when the present is on fire. When 3dfx sold most of its assets to NVIDIA at the end of 2000, it felt shocking but not entirely surprising. The company that had made PC gaming feel magical had run out of road.

What still matters

The reason people still talk about 3dfx is not just nostalgia, although nostalgia is definitely part of it. Old PC gamers can become misty-eyed over a driver disk if the lighting is right. But 3dfx matters because it changed expectations. It proved that ordinary consumers would pay for dedicated 3D hardware if the experience was dramatic enough. It helped create the idea that the graphics card was central to gaming, not an afterthought.

It also showed the power of developer relations. Hardware companies do not win only by building chips. They win by making developers care, by making games run well, and by giving consumers confidence that their expensive purchase will actually be used. That lesson remains central today.

The company’s failure is equally instructive. A strong brand can fade. A technical lead can disappear. A proprietary advantage can become a burden. Partners can become rivals. A bold acquisition can turn into an anchor. And in a fast market, being right too late can look almost the same as being wrong.

3dfx also left behind a more emotional legacy. It made PC gaming feel like a hobby with secrets, rituals, and rewards. Installing a Voodoo card was not as simple as buying a sealed console and plugging it in. It required effort. It required knowledge. It occasionally required prayer. But when it worked, the payoff felt personal. Your PC was not just running a game. It was revealing something.

The final frame

Today, 3dfx exists mostly in memory, collector forums, retro builds, old magazine scans, and the occasional dusty card pulled from a forgotten box in someone’s attic. Its technology has long been surpassed. A modern integrated graphics chip could probably outperform a Voodoo card while also checking email, decoding video, and silently judging your cable management.

But legacy is not only about performance. 3dfx arrived at the precise moment when PC gaming needed a leap, and it delivered one. It gave players smoother worlds, faster games, and a reason to believe that the PC was the most exciting gaming platform on earth. For a time, the Voodoo name meant you had the good stuff.

The company rose because it understood the future before others did. It fell because the future kept moving. That is the strange beauty of the 3dfx story. It was a pioneer, a champion, a cautionary tale, and a legend. It helped open the portal to modern PC gaming, then vanished through it. And for anyone who remembers that little 3dfx logo appearing before a game, the feeling remains hard to explain without sounding slightly ridiculous. But then again, that was always the point. It was Voodoo.